A lot of PDFs already have a text layer — running OCR again is just wasted work.
Firecrawl’s open-source pdf-inspector first checks the PDF type locally (in tens of milliseconds), then decides which files can be extracted directly and which pages actually need OCR.
Readable ones become Markdown; only the missing-text parts get OCR.
Their data: about 54% of PDFs don’t need OCR at all.
Perfect as a pre-check step for AI knowledge bases and document Q&A.
👉